yinghaohuang91/intercap

Source code for the GCPR 2022 paper: <InterCap: Joint Markerless 3D Tracking of Humans and Objects in Interaction>

intercap's Issues

Is there any way to directly download the dataset on linux server?

Code that would work on linux terminal would be great!

backprojection

Thank you for open sourcing this dataset!

Could you explain the calibaration files for peforming depth back-projection (depth to world space) ?
I found 3 different calibartion directories at: https://github.com/YinghaoHuang91/InterCap/tree/master/Data
wondering that which one is mapped to the recording 01/01/Seg_0/Frames_Cam1

local variable 'tmp' referenced before assignment

in obj_track/utils.py function load_targets,

for i in range(0, 6):
        tmp = np.asarray(tmp, dtype=np.float32) / 1e3
        tmp = torch.from_numpy(tmp)#.to(device)
        targets.append( tmp )

    return targets, bbs

local variable 'tmp' referenced before assignment

question about folder structure

Thanks for your great work!
In "readme.md", you said "Sample folder structure shown in: Data/recordings/subject01_motion01_seg01/Depth_X", however, I could not find "recordings" in "Data", could you please check for this? Thank you very much!

Code for preprocessing data

Hello,

Thanks for your work. I've noticed detailed explanations in the readme regarding the models used for preprocessing the data. However, I'm curious if there are any corresponding code references within the repository.

Specifically, I'm particularly interested in the calculation of overlapping areas between segmented objects in consecutive frames during object segmentation. I was wondering if there is any code related to this within this repository? Alternatively, do you have any other work or references that might be helpful in this regard?

I eagerly await your response.

你好，关于项目的运行麻烦能做个中文的说明吗

Could you provide annotation of smplx

Hi, I could not find annotation of smplx. Could you provide annotation of smplx? Thanks!

Code for rendering

Dear authors,

thank you for your amazing work!

We're interested in rendering only the object in order to create a ground-truth object mask for each frame. Would it be possible to share the script used to generate the rendered human-object mesh images?

Thank you very much in advance!